Motion Vector Prediction for Video Coding Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression algorithms are inefficient in encoding motion vectors, particularly in the presence of horizontal and vertical motion boundaries, as they often encode non-zero differences multiple times or fail to efficiently handle zero differences, leading to suboptimal bit usage.
Innovation Solution
A method and apparatus for encoding and decoding motion vectors using multiple predictions, where the encoder selects one prediction from multiple options based on context and encodes the difference, allowing for efficient bit allocation by reducing the number of bits required to code motion vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single prediction method (median operator or single reference block) is used for encoding motion vectors, then the encoding process is simple, but the bit efficiency deteriorates in the presence of horizontal or vertical motion boundaries
Solution Approach 1:
The prediction process is segmented into multiple independent prediction methods (first prediction from top block, second prediction from left block). Each prediction method operates independently and can be selected based on the local motion characteristics, allowing the system to adapt to different boundary orientations without increasing overall complexity
Solution Approach 2:
The prediction method dynamically adapts to the local motion characteristics by selecting between multiple prediction strategies based on the context. The system transitions from a static single-prediction approach to a dynamic multi-prediction selection approach that responds to horizontal and vertical motion boundaries
2Ease of manufacture
If the median operator is used for prediction, then the encoding is straightforward, but non-zero differences are encoded multiple times in the presence of horizontal motion boundaries
Solution Approach 1:
The prediction approach is segmented into direction-specific predictions: vertical prediction (from top block) for horizontal boundaries and horizontal prediction (from left block) for vertical boundaries. This segmentation allows each prediction to be optimized for its specific boundary type, reducing redundant encoding
Solution Approach 2:
Different prediction strategies are applied locally based on the boundary orientation. For horizontal boundaries, vertical prediction from the top block is used; for vertical boundaries, horizontal prediction from the left block is used. This local adaptation minimizes non-zero differences in each specific context
3Loss of information
If the single reference block prediction is used, then zero differences are generated for vertical motion fields, but the prediction becomes inefficient in the presence of vertical motion boundaries
Solution Approach 1:
The prediction system achieves multi-functionality by incorporating multiple prediction methods that can handle different boundary types. The first prediction handles vertical motion fields and horizontal boundaries, while the second prediction handles horizontal motion fields and vertical boundaries, creating a universal prediction system
Solution Approach 2:
The prediction parameters are changed based on the boundary type detected in the local region. The system switches between different reference blocks (top block vs. left block) and different prediction directions based on the motion field characteristics, optimizing the prediction for each specific scenario
Data Source
AI summary
Video compression algorithms typically represent visual information by a combination of motion and texture data. Motion data describes the temporal relationship between the content of a frame and that of a previous frame. This invention describes a method and apparatus for efficiently encoding motion data, particularly in the presence of horizontal and/or vertical motion boundaries.